Research On Grid Resource Management And Scheduling Strategies

Grid computing is a new type of distributed computing, and its goal is to achieve a wide range of resource sharing. Researches on resource management and scheduling Strategies have become one of the focuses in grid systems as its multiple characteristics like distributed, dynamic, diversity, autonomy and management. Most of the current grid resource management and scheduling systems in existence are poor on the adaptability and quality of service as well as weak support in combined resource allocation problem. Therefore, further research on the grid resource management and scheduling problems will have important theory significance and the practice value for the current and future development of grid and the application.Resource management problem is one of the key issues of electric power information grid to provide high-performance services. Based on the insufficient of existing models of grid resource management and the characteristics strategies of the power grid information system, this paper put forward a kind agent-based grid resource management model which is suitable for the power information grid. This paper divided the grid into grid domain and given a grid resource management system structure and domain of agent management structure named "center management-the coordination of local distribution" based on agent. This paper also provided definition of various agent levels and functions, and given multi-agent cooperation to achieve a complete work process for grid system resource management. This model has characters of highly efficient management, traffic load balancing, intelligent features, and practical.In order to ensure the service quality of grid, how to design a grid job scheduling strategy,which can adapt to a variety of operational characteristics and needs of different users, has also become an important research direction. According to the needs of users and the features of grid job scheduling, this paper presents a method for local grid system based on the grid job scheduling strategy of multi-priority preemption. In this strategy, grid job requests are divided into three categories:best-effort job requests, non-VIP users requests and VIP users requests. Different job types have different priorities. General requests from the VIP users have been given reserved resources, while general grid job request is given the priority to seize the best-effort job request. The simulation result shows that while quality of service of the jobs from non-VIP users and the best-effort jobs affect is relatively small, the strategy can fully guarantee the service quality of grid system job request from the VIP users.In addition, the strategy can reserved resources automatically to get the optimal performance service of grid system, according to the general operation strength and ratio change between general request and best-effort.